Should Everyone Take Metformin?

Gregory A. Nichols, PhD


December 04, 2014

In This Article

Analysis and Commentary

Randomized controlled trials are the hallmark of scientific investigation. However, they cannot answer all of the important questions about relative advantages of one drug over another. Drug registration trials are too short to recognize even medium-term effects on clinical outcomes.

Although longer trials are now being required for antihyperglycemics to demonstrate cardiovascular safety and benefit, such studies may be too controlled to account for the multitude of potential variables that more accurately reflect how new agents will be used in clinical practice. Thus, there is a clear need for comparative effectiveness studies that capitalize on the rich data sources now available from electronic medical records. To be useful, however, the studies must be well designed. The two articles reviewed here are excellent examples.

Berkowitz and colleagues pitted four classes of antihyperglycemics head to head, a comparison that would have been enormously expensive to conduct as a randomized trial. The use of time to therapy intensification was reminiscent of the ADOPT trial,[1] but two classes of drugs were included that were not yet available when ADOPT was conducted. The follow-up may have been too short to say much about secondary outcomes, but metformin demonstrated clear superiority over SUs, TZDs, and DPP4s with respect to time to treatment intensification.

The major shortcoming of the analysis was the unavailability of A1c values. Lower A1c before initiation of anti-hyperglycemic therapy has been shown to be an important predictor of attaining glycemic control and of the durability of maintaining it.[2,3,4,5] If there were substantial differences in A1c between the four patient groups at time of initiation, this would contribute to the time to intensification. In particular, patients initiating the newer DPP4 drugs may have had poorer control at baseline, a classic case of confounding by indication. Nevertheless, the differences between metformin and the other drugs were substantial, and although correction for preinitiation A1c could attenuate the reported results, it would be unlikely to eliminate them altogether.

Treatment intensification after the first anti-hyperglycemic sits at one end of the spectrum of diabetes outcomes, whereas mortality (as used by Bannister and colleagues) is the "final" outcome in epidemiologic analysis. The novel approach used in this study—an approach that would probably never be a part of a clinical trial—was the comparison with matched cohorts who were not exposed to the drugs being compared, and who did not have the condition for which the drugs were prescribed.

The protective effect of metformin relative to SUs was not a surprise; this was reported in the United Kingdom Prospective Diabetes Study as well as subsequent epidemiologic analyses.[6,7,8,9] The comparison of SU users with their matched controls was also expected, because mortality rates have long been reported to be approximately double among those with diabetes vs nondiabetics.[10] The remarkable finding was an apparent protective effect of metformin compared with unexposed nondiabetic individuals.

Because of metformin's favorable results among people with diabetes, it has been postulated that the drug may also provide benefit to people without diabetes. However, in one recent randomized controlled trial in patients without diabetes, progression of mean distal carotid intima/media thickness did not differ between patients allocated to metformin vs placebo during an 18-month follow-up.[11] Yet this was not a perfect endpoint for determining whether metformin could provide a mortality benefit in people without diabetes.

Whether metformin does indeed reduce mortality will probably not be answered with a long-term randomized controlled trial, nor are head-to-head studies likely to be conducted to determine the optimal order of anti-hyperglycemic therapies as the number of options continues to grow. Nonetheless, there is much to learn from meticulously conducted comparative effectiveness studies such as these.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.